Gerrymandering Opposition:
Minority-Concentrated Districts and Electoral Competition in Mexico
Anne Meng
Travers Department of Political Science
University of California, Berkeley
Brian Palmer-Rubin1
Travers Department of Political Science
University of California, Berkeley
[email protected]
858-335-5257
June 29, 2015
Word Count: 9,812
1
Corresponding author
Gerrymandering Opposition
2
Gerrymandering Opposition:
Minority-Concentrated Districts and Electoral Competition in Mexico
Abstract
Can institutions that are designed to improve minority representation also have an
effect on electoral competition? We address this question by examining how minorityconcentrated districts (MCDs)—designed to empower indigenous populations—affected
minority participation and party competition in Mexico. Using an original dataset and a
matching design that helps alleviate causal inference problems inherent to observational
studies, we find that MCDs had no effect on minority participation but enhanced electoral
competition. Field-research reveals that MCDs weakened one-party dominance by
assembling minority voting blocs that were amenable to opposition-party appeals. More
broadly, our results suggest that the mobilization of minority voting blocs can promote
electoral competition in transitional democracies.
Keywords: redistricting; minority participation; electoral competition; indigenous
politics; Mexico
Gerrymandering Opposition
3
Gerrymandering Opposition:
Minority-Concentrated Districts and Electoral Competition
Multi-ethnic societies undergoing democratization often face two challenges:
representing minorities and promoting electoral competition. States commonly address
minority underrepresentation by adopting institutions such as minority-concentrated
electoral districts, legislative quotas, and proportional representation. However, whether
such institutions achieve their desired goals remains contested. Furthermore, establishing
effective multi-party competition often presents the additional challenge of dismantling
dominant-party strongholds inherited from the previous regime. Many countries that hold
regular elections still suffer from one-party dominance, either on the national level or in
some subnational units. Can institutions that are designed to deal with the problem of
minority underrepresentation also have an effect on electoral competition?
In this paper, we address this question by analyzing the effects of indigenousconcentrated congressional districts in Mexico, a country that experienced an extended
period of one-party dominance. These districts, which we call minority-concentrated
districts (MCDs)2, were created in 2005 to encourage the political participation of the
historically marginalized indigenous population, our minority group of interest. We
observe the effect of these districts on the electoral participation of indigenous voters
moved into such districts and examine how this reform affected electoral competition in
2
We use the term minority-concentrated districts (MCDs) rather than the more
conventional majority-minority districts because we adopt the Mexican government’s
threshold for such a district, which is that at least 40 percent of its inhabitants are ethnic
minorities.
Gerrymandering Opposition
4
redistricted municipalities. We find that these institutions had little effect on minority
participation, but that they enhanced multi-party competition in a surprising way by
allowing opposition parties to penetrate dominant-party strongholds.
This project implements a multi-method strategy, combining statistical
approaches that exploit variation induced by redistricting with field research findings
aimed at identifying mechanisms. Our quantitative identification strategy capitalizes on
an episode in Mexico in 2005 where congressional district boundaries were redrawn,
reassigning some highly indigenous municipalities to MCDs that concentrated indigenous
voters. We estimate the effect of this intervention on turnout, dominant-party vote share,
and vote margins. We also conduct process tracing using field-research findings to
illustrate the causal process connecting MCDs to minority voting behavior in Mexico.
This article contributes to the nascent literature on minority participation in
transitional democracies by measuring the effect of MCDs on turnout in Mexico.
Adapting a research design developed by Sekhon and Tituinuk (2012), we exploit
variation induced by redistricting and implement a matching design that helps alleviate
causal-inference problems inherent to observational data to estimate the effect of MCDs
on minority participation. We find no significant effect of MCDs on minority turnout. We
argue that there is a lack of an effect because members of Mexico’s indigenous minority
have historically been incorporated in the dominant party—albeit in a marginalized
way—and participated in elections at roughly the same rate as non-indigenous rural
populations prior to the reform.
We also extend beyond the typical approach to studying MCDs by examining
their effect on electoral competition and one-party dominance. Given that MCDs alter the
Gerrymandering Opposition
5
demographic composition of congressional districts, it stands to reason that they may
affect the competitive balance between parties that challenge for office. We redeploy the
above-described matching design to gauge the effect of MCDs on dominant-party vote
share and margins between the top two vote-getting parties in a municipality. If MCDs
enhance electoral competition, municipalities assigned to these districts should reflect a
lower vote share for the dominant party as well as lower vote margins. We find that
MCDs have a significant negative effect on dominant-party vote share and on vote
margins, suggesting that the creation of these districts bolstered electoral competition.
We use qualitative findings to show that the creation of congressional districts that
concentrated minority voters provided opposition parties with an opportunity to penetrate
dominant-party strongholds by mobilizing these new voting blocs.
This paper makes three important contributions. First, we evaluate whether a
policy intervention, designed to empower Mexico’s indigenous population, achieved one
of its normative goals: increasing indigenous voter turnout. Despite the prevalence of
minority-promoting institutions in new democracies, few empirical studies outside of
United States test the efficacy of such institutions. Our paper joins a nascent group of
scholarship in comparative politics that estimates the effect of MCDs, legislative quotas,
and proportional representation on minority representation. Second, we illustrate how
minority-promoting electoral institutions can introduce electoral competition by opening
up new space for opposition parties to compete with entrenched dominant parties—a
surprising consequence of MCDs. Finally, we make a methodological contribution to the
comparative politics literature by implementing a research design that exploits
redistricting to estimate the causal effect of MCDs.
Gerrymandering Opposition
6
The next section of this paper presents our argument of voter mobilization in
MCDs in light of existing scholarship. We then conduct an overview of indigenous
politics and party competition in Mexico. We test the effect of MCDs on voter turnout,
dominant-party vote share, and vote margins in the following section, exploiting a
redistricting initiative that occurred in 2005 as part of our identification strategy. We
present our data, research design, and empirical findings, as well as ecological-inference
analysis. We then present field-research findings from the state of Chiapas to provide
evidence that MCDs increased electoral competition and weakened the dominant party
through the mechanism of opposition-party appeals to minorities who had previously
voted overwhelmingly for the dominant party. We conclude by discussing the
implications of our study for broader debates of minority representation and electoral
competition in transitional democracies.
Voter Mobilization in Minority-Concentrated Districts
MCDs have been created in the United States and elsewhere to enhance minority
participation in electoral politics and representation in policymaking. It is contested
whether such institutions achieve these goals in the United States, and scholars are only
beginning to address their impacts in transitional democracies. Bobo and Gilliam (1990)
contend that having minority officeholders empowers minority constituents, yielding
greater participation. They argue that “(b)lacks in high empowerment areas [cities with
black mayors] should feel more trusting of government, express higher levels of efficacy,
and become more knowledgeable about politics than blacks in low-empowerment areas.
All of which should, in turn, contribute to higher levels of participation” (p. 379).
Gerrymandering Opposition
7
Subsequent studies of MCDs in the United States have found evidence in line with the
“minority empowerment theory”: these districts yield more minority officeholders and
greater minority turnout (Tate, 1991; Voss & Lublin, 2001; Barreto, Villarreal, & Woods,
2005; Barreto, 2007). On the other hand, Gay (2001) finds that the presence of minority
incumbents actually lowers white political participation, rather than enhancing minority
turnout. In a study of redistricting more broadly, Hayes and McKee (2009) contend that
the redrawing of district boundaries raises information costs for voters by severing ties
between incumbents and their constituents and thus can actually lower participation. In a
separate study, they find that this “disruptive” effect of redistricting disproportionately
affects minority voters, who face higher information costs than majority-group voters.
However, when minorities are redrawn into a minority candidate’s district, the effect is
reversed and redistricting enhances minority participation (Hayes & McKee, 2012).
MCDs purported effect on minority participation has also been challenged on
methodological grounds. Sekhon and Tituinuk (2012) show that observational studies
that compare MCDs with non-MCDs without taking into account how these districts are
drawn yield biased estimates. Henderson, Sekhon, and Titiunik (2012) and Keele and
White (2011) find no effect of MCDs on minority turnout after adjusting for selection
effects—namely that minorities who were moved into these districts were more
participatory than minorities who were left in non-MCDs. These findings suggest that
estimates of MCDs’ effects must control for political biases in the redistricting process.
Adapting Sekhon and Tituinik’s research design to deal with such biases, a first
goal of this paper is to contribute to the burgeoning group of scholars that analyze
minority-promoting electoral institutions outside of the United States. Only a handful of
Gerrymandering Opposition
8
existing studies examine MCDs or other minority-promoting institutions such as quotas
on minority participation in transitional democracies, and their findings have been mixed.
Goodnow and Moser (2012) find that majority-minority districts in Russia increase
minority turnout when a minority is on the ballot, in line with minority empowerment
theory. If this theory holds in our case, we should find that indigenous voters who have
been reassigned to MCDs turn out in larger numbers than indigenous voters who have
not. On the other hand, studies of legislative quotas for Scheduled Castes (SCs) in India
have found that turnout decreases in constituencies reserved for SC candidates, in part
because voters are not convinced that SC representatives would promote their interests as
well as previous non-SC representatives (Jensenius, 2013; McMillan, 2005). These mixed
findings suggest that the effect of minority-promoting electoral institutions on minority
participation is contingent on pre-existing alignments between ethnic minorities and
political parties. In the Mexican case we consider how political parties mobilized
indigenous populations prior to redistricting to understand how MCDs influenced
indigenous turnout. If minorities were historically mobilized effectively through
patronage appeals, then we would not expect MCDs to find a minority empowerment
effect.
Our second goal is to extend beyond the typical approach to the study of MCDs to
examine their effect on electoral competition in transitional democracies. Even after such
countries experience national transitions to multi-party competition, formerly dominant
parties often remain unchallenged in certain electoral races. For example, in many
countries that have democratized at the national level, “subnational authoritarian”
enclaves persist, where dominant parties manipulate state resources and erect barriers to
Gerrymandering Opposition
9
opposition-party entry (Gibson, 2012; Giraudy, 2010). Here we argue that MCDs—
institutions
designed
to
bolster
minority
participation—can
have
unexpected
consequences for electoral competition.
Research on the partisan effects of redistricting in American politics suggests that
the creation of MCDs can have a dampening effect on two-party competition (Brace,
Grofman, & Handley, 1987; Hill, 1995). This occurs because MCDs “pack” a limited
number of districts with minority voters who, in the United States, are expected to vote
for the Democratic Party. Doing so, however, comes at the expense of making other
districts less competitive for the Democratic Party because minority voters have been
siphoned off to MCDs.3 However, we expect the effect of MCDs on electoral competition
to differ in Mexico, where indigenous voters do not have the same bonds of loyalty and
programmatic congruence with a given party as minority voters have with the Democratic
party in the US. Rather, Mexico’s indigenous populations were incorporated into the
dominant party in the post-revolutionary period through a state corporatist system that
rewarded loyalty with patronage and punished dissent with repression and withdrawal of
financial resources (Rus 1994, Diaz-Cayeros, Magaloni, & Weingast 2003). Because of
this, when minority voting blocs are provided with a viable and appealing alternative to
the PRI, we would expect them to be amenable to party switching.
In fact, a number of existing studies have uncovered cases of opposition parties
that penetrated one-party dominance by courting previously disenfranchised or
electorally demobilized constituencies. Chhibber (1999), for example, shows how the
3
Cameron, Epstein, and O’Halloran (1996) point out that this concentration of minority
voters also reduces overall support for legislation that advances minority interests.
Gerrymandering Opposition
10
dominance of the Congress Party in India declined when opposition parties catalyzed a
cleavage between upper and lower castes. Scholars of the United States argue that the
1965 Voting Rights Act, which enfranchised black voters by prohibiting poll taxes,
literacy tests, and all-white primaries helped the Republican Party penetrate the once
Democrat-dominated South (Valelly, 2004; Mickey, 2008). Following a similar logic,
scholars of ethnic voting in new and transitional democracies have noted that minority
groups often form de facto voting blocs by strategically voting with their ethnic group,
often for co-ethnic candidates, whom they presume will reward them with favorable
policies or discretionary spending (Birnir, 2007; Chandra, 2004; Posner, 2005). Other
scholars have studied the electoral consequences of an institutional reform in Mexico—
the creation of municipal-level indigenous autonomy regimes (Usos y Costumbres) in
Oaxaca, a state notable for subnational authoritarianism (Benton, 2012; Cleary, 2009;
Eisenstadt, 2007). They find that this reform enhanced one-party dominance, in large part
because the dominant party manipulated the reform to empower its indigenous allies.
Building on these studies, we argue that the creation of MCDs can enhance
electoral competition and diminish one-party dominance by opening space for previously
non-competitive parties to build new constituencies. When minority voters are reassigned
from districts with low minority populations to MCDs, minority groups become newly
influential voting blocs in these districts. It follows that all political parties have a
heightened incentive to incorporate these groups into their coalitions given their
increased electoral weight. Parties may appeal to minority voters through targeted
patronage offers or the nomination of minority candidates for elected office. Minority
voters—if capable of acting collectively—are prone to throw their support behind the
Gerrymandering Opposition
11
party that most credibly promises to deliver such inducements. A sizeable and sudden
change in the ethnic composition of a district may enhance opposition parties’ ability to
make such promises to minority groups in two ways: First, dominant parties are often
constrained by commitments to entrenched majority-group constituencies that were
constructed during their period of rule while the lack of a deeply rooted constituency
affords opposition parties more flexibility to divert electoral inducements to minorities in
the aftermath of redistricting. Second, if the opposition party had already positioned itself
as the minority-friendly option prior to redistricting, the creation of MCDs may signal to
minority voters in such districts that the opposition party stands a stronger chance of
winning—and thus will control local budgets which can be used to deliver patronage.
Where opposition parties are successful at gaining a following among minority
voters in MCDs, elections will become more closely contested. We probe for this
association in the Mexican case by comparing dominant-party vote share and margins
between the top two vote-getting parties in congressional elections for municipalities in
MCDs and non-MCDs. If our argument holds, we should find that opposition parties
were more successful at attracting indigenous voters in MCDs following redistricting
than in non-MCDs, yielding more competitive elections.
This hypothesized effect of MCDs on electoral competition can have a limited
duration, as the dominant party may eventually break ties with non-minorities and make
appeals to indigenous voters that are competitive with opposition-party appeals. Even if
the once-dominant party does “catch up” in this way, however, the effect of MCDs
remains meaningful, as it has created conditions propitious for opposition parties to
weaken dominant-party strongholds.
Gerrymandering Opposition
12
Indigenous Politics and Party Competition in Mexico
The Mexican case offers a propitious context to observe the electoral effects of
MCDs owing to the presence of a politicized ethnic minority and a history of dominantparty rule. Like most Latin American countries, the bulk of Mexico’s population belongs
to two broad ethnic categories: indigenous and mestizo. Mexico’s indigenous population,
our minority group of interest, constitutes roughly ten million people, making up 9.5
percent of the national population and is the largest indigenous population in any Latin
American country. Although existing definitions of indigeneity identify multiple
dimensions, we adopt a common proxy based on whether citizens report speaking an
indigenous language in the 2005 census. The dominant ethnic group is mestizo, which we
define as including all Mexicans who do not report speaking an indigenous language. The
indigenous population in Mexico is composed of 62 distinct ethno-linguistic groups, the
majority concentrated in the southeastern region of the country. Fifty-eight percent of
Mexico’s indigenous live in the five states of Oaxaca, Chiapas, Veracruz, Yucatan and
Puebla.4 However, sizeable indigenous populations (at least 50,000) appear in 23 of
Mexico’s 32 states. The indigenous tend to live in poor rural areas and experience
significantly lower levels of almost all social-development indicators than the general
Mexican population (UNDP, 2010). 79.3 percent of Mexico’s indigenous live in poverty
4
Data taken from Mexico’s National Commission for the Development of Indigenous
Peoples. Retrieved from:
<http://www.cdi.gob.mx/index.php?option=com_content&view=article&id=758&Itemid
=68>
Gerrymandering Opposition
13
and 40.2 percent live in extreme poverty, compared to 44.1 percent and 8.5 percent for
the general Mexican population (CONEVAL, 2010).
Indigenous populations have also historically been marginalized from political
power, but mobilized electorally in favor of the dominant party through patron-client
relationships with local mestizo elites. Beginning in 1936, several indigenous
communities were incorporated into the corporatist structure of the ruling party through
the Sindicato de Trabajadores Indígenas (Indigenous Workers’ Union), granting these
populations access to land and financial benefits in exchange for political quiescence
(Rus, 1994). Throughout the twentieth century, indigenous authorities within the
dominant Partido Revolucionario Institucional (Institutional Revolutionary Party, PRI)
were subordinate to the party’s mestizo-dominated sectors, such as the Confederación
Nacional Campesina (National Peasant Confederation, CNC), assuring that indigenous
figures rarely occupied important leadership posts within the party or were nominated for
elected office. Nonetheless, clientelist mobilization has historically turned out indigenous
voters in large numbers for the PRI. In fact, municipal-level ecological inference
estimates suggest that indigenous voters were turning out in higher percentages than
mestizo voters prior to the 2005 redistricting—an average of 44 percent indigenous
turnout versus 42 percent mestizo turnout in the 2003 legislative election and 49 percent
indigenous turnout versus 40 percent mestizo turnout in the 1997 election.
Despite this broad support for the PRI, important instances of indigenous
rebellion began to emerge in southern Mexico in the last decades of the twentieth
century. Non-state actors—including Catholic and Protestant missionaries—helped
construct networks between diverse indigenous communities and create a pan-indigenous
Gerrymandering Opposition
14
identity that found its expression in a series of uprisings, most famously the 1994
Zapatista movement in Chiapas (Harvey, 1998; Trejo, 2009). While the Zapatista
movement conspicuously spurned electoral politics, other contentious indigenous
organizations formed party alliances in the aftermath of widespread protest.
Mexico has a notable legacy of dominant-party rule by the PRI, though the 2005
redistricting reform took place during a period of national-level decline for the PRI,
including the loss of the presidency to the Partido Acción Nacional (National Action
Party, PAN) in 2000. Having held the governorship of all 32 states since its inception in
1929 until 1989, by 2005 the PRI had lost at least one gubernatorial election in 15 states
to the PAN or the Partido de la Revolución Democrática (Party of the Democratic
Revolution, PRD), the two main opposition parties. Furthermore, the congressional
period during which the redistricting took place (2003-2006) was the first time since its
inception that the PRI did not control the presidency, nor held a majority of seats in either
congressional chamber.
Despite this national-level slump, the PRI retained a significant advantage in
electoral politics in many subnational units, particularly in predominantly indigenous
regions. For example, in the 2003 congressional election, the PRI won 22 out of 23
possible seats in Chiapas and Oaxaca, the two states with the largest indigenous
populations. The PRI enjoyed large vote advantages among indigenous populations,
owing to long-standing patron-client relationships with these groups and cooptation of
local indigenous authorities and the PAN and PRD’s lack of party infrastructure in the
countryside. Given the upsurge in indigenous activism beginning in 1994, corresponding
with national trends of increased multi-party competition, opposition parties saw an
Gerrymandering Opposition
15
opportunity to break the PRI’s stranglehold in highly indigenous states by courting
indigenous voters. They were largely successful; of the 23 districts in Chiapas and
Oaxaca, the PRI only won nine in 2006 and 13 in 2009.
Statistical Analysis: Redistricting and Minority Voting
To identify the effect of MCDs on minority turnout, dominant-party vote share,
and electoral competition, we capitalize on a 2005 redistricting initiative that reassigned
some municipalities to MCDs. The creation of new MCDs—defined as districts with
indigenous populations greater than 40 percent—was an overtly stated goal of the
redistricting process.
We constructed an original dataset with electoral, demographic, and
socioeconomic data for all Mexican municipalities in the 19 states that have at least one
municipality that is majority indigenous. 5 Following previous studies that use
redistricting to measure electoral outcomes (e.g. Ansolabehere, Snyder, & Stewart 2000;
Keele & White 2011; Sekhon & Titiunik, 2012), we use the smallest unit of analysis for
which we have electoral and socio-economic data, permitting us to compare sub-district
units that are as similar as possible. Each electoral district is composed of multiple
municipalities, and redistricting involves municipalities being reassigned from one
district to another. 6 Through Mexico’s Federal Electoral Institute we acquired each
5
These include all eleven states with MCDs plus eight states that do not have MCDs. The
resulting dataset has 2,048 municipalities out of a total of 2,457 in Mexico.
6
Given that the average population of districts is over 325,000 inhabitants, only highly
urban municipalities compose their own districts or are split into multiple districts. None
Gerrymandering Opposition
16
municipality’s old (pre-2005 redistricting) and new (post-2005 redistricting) district
numbers and calculated the percentage of the populations that are indigenous in each
municipality’s “old” and “new” districts. We also collected municipal-level electoral
returns for the 1997, 2003, and 2009 midterm elections of Mexico’s lower house, the
Chamber of Deputies.7 This chamber elects representatives every three years, without
reelection, from 300 single-member federal districts, split among 31 states plus Mexico
City.8
From this information, we were able to calculate three dependent variables on the
municipal level, our unit of analysis: 1) voter turnout, 2) vote share for the PRI, and 3)
vote margin between the two highest vote-getting parties. PRI vote share can be
interpreted as measuring the degree of one-party dominance in the municipality, since the
PRI is the historically dominant party for all municipalities. We are interested in seeing
whether PRI vote share was affected by MCDs, even in cases where the PRI remained the
of these urban municipalities have sizable indigenous populations and these are thus
excluded from the analysis.
7
We chose to analyze midterm elections for two reasons: first, the 2009 midterm election
was four years after the 2005 redistricting, allowing more time for this treatment to take
hold than the 2006 election; second, we expect midterm elections to better reflect the
response of voters to their districts’ composition than presidential elections, where
coattail effects of the presidential race may play a larger role.
8
The prohibition of reelection in Mexico lessens the risk that our results are confounded
by an incumbency effect of the type described by Hayes and McKee (2009).
Gerrymandering Opposition
17
highest vote getter following redistricting.9 For instance, if the PRI received the most
votes in both 2003 and 2009 in a particular municipality but received a significantly
lower vote share in 2009, we would interpret this decline in vote share as reduced oneparty dominance in the municipality. In addition, we use the vote margin between the top
two vote-getting parties in the municipality as our measure of electoral competition.
We merged this electoral data with demographic and socioeconomic data from
Mexico’s 2005 census, including municipality indigenous population, municipal budgets,
population, and several household measures of socioeconomic status. Summary statistics
are presented in Appendix A.
We use a variation of the research design developed by Sekhon and Titiunik
(2012) and Henderson, Sekhon, and Tituinik (2012) by comparing municipalities that
were reassigned to MCDs during redistricting to similar municipalities that were not. We
define a treatment municipality as one that was reassigned to a new district that is at least
50 percent more indigenous than its old district and whose new district is at least 40
percent indigenous. We chose to define treatment in this way, rather than with an absolute
threshold for an MCD (such as a district that is at least 40 percent indigenous) because
we are interested in the effects of a sizable change in district composition rather than
simply the effect of being in a highly indigenous district. Thus municipalities that were
already in MCDs prior to redistricting do not constitute treatment units. Control units are
9
In our main dataset, the PRI received the most votes in 88 percent of the municipalities
in 2003 and 85 percent of the municipalities in 2009. In 2003, the PRI received the
second-most votes in all municipalities for which it was not the top vote getter and only
failed to receive the first or second-most votes in two municipalities in 2009.
Gerrymandering Opposition
18
municipalities residing in districts that did not experience a 50 percent increase in
indigenous populations or are in districts that are less than 40 percent indigenous after
redistricting. Figure 1 provides a graphical representation of treatment assignment.
Triangles represent treatment municipalities and circles represent control municipalities.
Figure 1
Treatment Assignment
0.8
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●●
●
● ●
●
●● ●
●
●
●
● ●
●
Percent Indigenous of 2009 District
●
●
●●
●●
●
●●
●
●●
●
●
●●●●
●●
●
●●
●
●
●
●
●●
●●
0.6
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●●
●●
●●
●●●
●
●
●
●
●●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●●●
●●●●
●●●●
●
●
●
●
●
●●●
●
●
●●
●
●
●●
●●●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●
●●
●
● ●●
●
●●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●
●●
●
●
●
●
●
●
●
●●
0.4
●●
●
●
●
●
●
●
●●●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
0.2
0.0
●
●
●●●
●
●
●
●
●
●●
● ●●
●
●●
●●
●
●
● ●
●
●●● ●
●
● ●
●
● ●●●●● ●
● ●●
●
●●
●●
● ●●
●
●
●●
●
●
●●
●
●●●
●
●●● ●
●
●
●
●●
●● ●●
●
●
●●
●
●●●
●●
●
●
●
●●
●
●
●●
●
●●
●
●
●
●
●
●
●
●●
●
●
●●
●●
●●
●
●
●
●
●●
●●
●●
●
●
●
●
●
●
●
●
●● ● ●●
●
●
●●
●●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●●
●
●
●●●
●
●
●
●●
●●
●
●●
●
●●
●
●
●
●
●
●
●
●
●● ●
●●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●●
●
●
●
●
●●
●●
●●
●●●
●
●
●
●● ●
●
●
●
●●
●
●
●●
●
●●
●
●●
●
●
●
● ●●
●
●
●●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
● ●
●
● ●
●
●
●●●
●●
●
●
●
●
●
●
● ●●● ●●
● ●
●
●
●●●
●
●● ●●
● ●●●
●
●● ●
●
●
●●
● ●
●
●●●
●●●
●
●
●●●●
●
●
● ●●
●
●
●
●
●
●●
●● ●
●
●●●
●
●●
●●
● ●●
●
●
●
●●
●
●●
●
●●●
●
●
● ●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●●
●●● ●● ● ●●
●●
●
●
●●●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●●
●
●
●●●
●●
●
●
●●●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
● ●●●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
● ● ●
●
●
●
●
●
●
●
●
●
●
●
● ●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
●●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
● ●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●● ●●
●
●
●
●
●●
●
●●
●
● ●●
●
●
●
●●●●●
●
●
●●●
●● ●
●●
●● ●●
●
●
0.0
0.2
●
●●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
● ●
●
●
●●
●
●●
●
●●
●●
●
●●
●
●
●
●
●
●
●●●
●
● ●
●
●
●
0.4
0.6
Percent Indigenous of 2003 District
Condition
Treatment
●
Control
Note: N = 2,048. “Jittering” was used to display each municipality as a distinct point.
Redistricting allows us to exploit both over-time variation and cross-sectional
variation by comparing electoral outcomes for treatment and control municipalities
before and after redistricting. The maps in Figure 2 indicate example treatment and
control municipalities in Chiapas pre- and post-redistricting. The municipality of Chanal
Gerrymandering Opposition
19
constitutes a treatment unit because it was reassigned from an old district that was 13
percent indigenous to a new district that is 70 percent indigenous. The municipality of
Huixtan constitutes a control unit because the composition of its district was unchanged
by redistricting.
Figure 2
Example of Redistricting in Chiapas
Before Redistricting
After Redistricting
Note: Bold lines represent district boundaries and dotted lines represent municipal boundaries.
A comparison of all treatment and control municipalities reveals that redistricting
did not constitute random treatment assignment. For instance, treatment municipalities
are poorer and supported the dominant party more than control municipalities prior to
redistricting. Comparing treatment municipalities to control municipalities without taking
these imbalances into account would potentially result in biased estimates of the
treatment effect of MCDs. To address this threat to inference, we matched treatment units
to control units that were most similar on variables that were potentially endogenous to
the redistricting process. To identify these variables, we conducted research on the
political biases of redistricting.
The 2005 redistricting reform (Federal Electoral Institute Agreement CG28/2005)
delineated ten criteria for the composition of new districts; the criterion central to this
Gerrymandering Opposition
20
study is that “the territorial integrity of indigenous communities will be preserved.” This
criterion was adopted in response to the Mexican Constitution’s mandate to “guarantee
the validity of indigenous rights and the integral development of their pueblos and
communities” and that indigenous communities would be concentrated in order to
“promote (indigenous) political participation” (IFE, 2005a: 13). Following redistricting
28 out of the 300 single-member districts in Mexico were greater than 40 percent
indigenous, compared with 17 prior to redistricting (IFE, 2005b: 101; Sonnleitner 2012b:
32-33). MCDs appear in 11 different states and encompass 494 municipalities.
Based on the ten redistricting criteria, the IFE applied a mathematical algorithm to
generate a “first redistricting scenario” using demographic data from the 2000 census and
submitted this scenario for review to a committee made up of six technical experts and 19
party representatives (IFE, 2005b: 11-17). The members of this committee were
nominated by congress and included representatives of every major party. 10 This
committee submitted 200 requests for specific changes to district composition, 70 of
which were adopted to derive the final district compositions.
Importantly, the redistricting reform was passed and implemented during the first
term for which the PRI did not hold the presidency, having lost to the PAN in 2000. The
PRI also did not hold a majority of seats in either congressional chamber, despite
10
Three of these 19 party representatives were appointed by the PRI, and two were
appointed by the Partido Verde Ecologista de México (PVEM), a party that often runs in
coalition with the PRI. The remaining 14 seats were split among parties that typically
oppose the PRI, including three from the PAN, four from the PRD, four from the Partido
de Trabajo (PT), and three from Convergencia (IFE, 2005b: 136-367).
Gerrymandering Opposition
21
retaining its dominance over several subnational regions. Since the PRI and its coalition
partner held only five of the 19 seats on the redistricting committee, the PRI was not in
position to control redistricting outcomes. It is plausible that the opposition parties—the
PAN, PRD and smaller parties that often run in coalition with the PRD—colluded within
the redistricting committee to modify districts in a way that would undermine PRI
strongholds.
Incorporating these insights about the redistricting process, we used Genetic
Matching, a non-parametric matching technique, to match each treatment municipality to
a control municipality that represents its best possible counterfactual, thus excluding
control municipalities that are significantly different from treatment municipalities.11 The
variables we included in the matching algorithm account for three main threats to
inference that may have resulted from the redistricting process.12 First, because MCD
creation disproportionately targets indigenous municipalities, treatment units are
systemically more indigenous, poor, and rural than control units. Therefore, we matched
on percent indigenous, logged municipal public spending, percent of households with
concrete floors (a proxy for poverty), and population. We also assured that the old
districts for the treatment and control municipality in each matched pair had a similar
11
Genetic Matching is an evolutionary search algorithm that achieves optimal “balance”
on covariates between treatment and control units using observational data. It has been
shown to outperform other matching techniques such as propensity-score matching
(Sekhon, 2011).
12
See Appendix B for a complete list of formal redistricting criteria. Several of these
were unlikely to introduce bias and thus not included in the matching algorithm.
Gerrymandering Opposition
22
percentage of indigenous people. Second, party representatives in the redistricting
committee may have targeted municipalities that were systematically different in their
political participation or support for the PRI. To control for these political biases, we
matched on PRI vote share and turnout in the 2003 congressional elections. (We did not
match on 1997 electoral data to preserve these variables for placebo tests.)
Finally, in order to control for the disruptive nature of redistricting, we matched
treatment and control variables on a measure of district turnover. While the prohibition of
reelection nullifies incumbent-specific information costs of reassignment, redistricting
can disrupt linkages between constituents and local party organizations or patronage
networks, thus raising non-incumbent specific information costs. To account for this
potentially confounding effect of MCD creation, we created an additional variable,
Percent New Neighbors, calculated as the percentage of the population from a given
municipality’s new district (“new neighbors”) that was not in its old district. Hence,
municipalities with higher values for this variable experienced more disruption in their
district compositions than municipalities with lower values. By matching treatment and
control units on Percent New Neighbors, we control for the disruptive effects of
redistricting, allowing us to isolate the effect of changes in districts’ ethnic composition
on electoral outcomes.
Genetic Matching produced a matched dataset of 246 municipalities, evenly split
between treatment and control units.13 (See Appendix C for more detail on the matching
13
These 246 municipalities exclude 12 treatment units that were dropped through the use
of a caliper because satisfactory matching control units could not be found. The caliper
Gerrymandering Opposition
23
process and the balance plot reflecting pre- and post-matching balance on covariates.)
Because we are most interested in the effect of MCDs on indigenous voting behavior, we
also analyze a subset of the full matched dataset, which we refer to as the majority
indigenous matched dataset. This dataset, which includes 88 municipalities, is composed
of all treatment municipalities that are at least 50 percent indigenous with their
corresponding control municipalities. (See Appendix D for descriptions of each dataset.)
Statistical Findings
We find no significant effect of MCDs on turnout and a significant negative effect
on dominant-party vote share and vote margins in highly indigenous municipalities. After
using Genetic Matching to match treatment municipalities to comparable control
municipalities, we conducted difference-of-means tests across matched pairs. These tests
yielded significant negative estimates of MCD effects for vote margins across both
matched datasets. Further, MCDs had a negative and significant effect on PRI vote share
in the majority indigenous matched dataset. MCDs appear to have no effect on turnout in
either dataset, but may have depressed mestizo turnout. Second, we conducted differencein-differences (DID) analyses on both matched datasets. The DID produced similar
results as the difference-of-means test.
We ran placebo tests for each of these models, testing whether treatment was
correlated with voter turnout, PRI vote share, or vote margin in 1997. If matching affords
conditional independence of treatment and outcomes, our treatment, which occurred in
was implemented to improve balance on municipal percent indigenous, the percentage
indigenous of the old district, and percent new neighbors.
Gerrymandering Opposition
24
2005, should not have any so-called “effect” on 1997 outcomes. Placebo tests on pretreatment outcome data allow us to check that observed differences between treatment
and control units are not attributable to pre-treatment heterogeneity. Results of the
placebo tests on turnout, PRI vote share, and vote margin were not significant.
Table 1 presents results from difference-of-means tests using both matched
datasets. We find no apparent effect of MCDs on voter turnout. In both datasets we find a
significant negative effect of MCDs on vote margins. The magnitude of this effect is
more sizable when analysis is restricted to majority indigenous municipalities. Finally,
being reassigned to an MCD appears to have a significant negative effect on the PRI’s
vote share in the majority indigenous matched dataset. Taken together, these results
provide evidence that reassignment to MCDs increased electoral competition for these
highly indigenous municipalities. 14
14
As a robustness check, we tested the effect of MCDs on the effective number of
parties, using the Laakso-Taagepera (1979) measure. The estimated effect was positive,
but not significant, likely due to the fact that the congressional elections observed
continued to be largely two-party races, albeit with narrower vote margins.
Gerrymandering Opposition
Table 1
Difference-of-Means Estimates
Outcome: voter
turnout
1997
Placebo
2009
Test
Matched
-.0150
-.0050
Dataset
(.0109)
(.0103)
(N = 246)
Majority
Indigenous
-.0171
.0162
Matched
(.0191)
(.0161)
Dataset
(N = 88)
25
Outcome: PRI vote
share
1997
Placebo
2009
Test
Outcome: vote
margin
1997
Placebo
2009
Test
-.0093
(.0127)
-.0062
(.0147)
-.0119
(.0214)
-.0530**
(.0178)
-.0144
(.0198)
-.0685*
(.0198)
-.0037
(.0293)
-.0894*
(.0248)
Note: * p<0.05, ** p<0.01. Standard errors are in parentheses.
The average difference in PRI vote share between matched treatment and control
municipalities in the majority indigenous matched dataset is 6.9 percentage points and the
difference in vote margins is 8.9 percentage points. The magnitude of the effect is
sizable: The average 2009 vote margin in this dataset was 24 percentage points and the
margin was less than 8.9 percentage points in 18 percent of the municipalities in our
sample. These results suggest that opposition parties were significantly more successful
at recruiting indigenous voters in municipalities that were reassigned to MCDs.
Moreover, the lack of a significant effect of MCDs on turnout indicates that the increased
returns for opposition parties are more likely a result of indigenous voters switching away
from the PRI rather than turning out to vote in larger numbers.
These findings are robust to different definitions of treatment. Using the same
matched dataset, we estimated the effect of two additional versions of treatment: (1) a
“magnitude” effect where treatment municipalities are defined as those whose new
districts are at least 50 percent more indigenous than their old districts; and (2) a
Gerrymandering Opposition
26
“threshold” effect where treatment municipalities are defined as those whose old district
was less than 40 percent indigenous and whose new district is greater than 40 percent
indigenous. Through both of these specifications, the estimated effect of MCDs on
turnout remains insignificant and estimates for PRI vote share and vote margin remain
negative and significant with similar point estimates. (Results reported in Appendix E.)
We also performed difference-in-differences (DID) tests on the two matched
datasets. Our DID estimator compares the difference between pre-treatment and posttreatment outcomes for each treatment unit against the difference between pre-treatment
and post-treatment outcomes for each control unit. These results, reported in Table 2,
produced similar results as the difference-in-means tests. Using the same binary treatment
variable as the difference-in-means analysis, we continue to find no significant effect of
MCDs on turnout, a negative effect on vote share in both datasets, and a negative effect
on PRI vote share in the majority indigenous matched dataset. The DID model on the
majority indigenous matched dataset produced significant negative estimates of 5.9
percent for PRI vote share and 8.3 percent for turnout. The magnitudes of these estimated
effects are similar to the difference-of-means estimates. We also ran placebo tests for the
DID models, which produced no significant effect of treatment on the difference between
1997 and 2003 outcomes. As a robustness check, we reran all DID tests using a
continuous measure of treatment—the change in the percentage indigenous of a
municipality’s district—as the independent variable. These tests also yielded negative
estimates for PRI vote share and vote margin and no significant estimate for turnout.
Gerrymandering Opposition
27
Table 2
Difference-In-Differences Estimates
Outcome: voter
turnout
19972003
2003Placebo
2009
Test
Full Matched Dataset (N = 246)
Outcome: PRI vote
share
19972003
2003Placebo
2009
Test
Outcome: vote
margin
19972003
2003Placebo
2009
Test
0.0010
(0.0149)
-0.0021
(0.0160)
0. 0007
(0.0230)
-0. 0419
(0.0214)
Continuous
0.0135
-0.0116
-0.0168
Treatment
(0.0517)
(0.0226)
(0.0305)
Majority Indigenous Matched Dataset (N = 88)
-0.0254
(0.0328)
-0.0092
(0.0472)
-0.0883*
(0.0439)
Main
Treatment
0.0194
(0.0213)
-0.0139
(0.0181)
0.0053
(0.0283)
-0.0594*
(0.0269)
-0.0029
(0.0408)
-0.0827*
(0.0359)
Continuous
Treatment
0.0314
(0. 0471)
-0.0553
(0.0394)
0.0058
(0.0623)
-0.1771**
(0.0577)
-0.0156
(0.0897)
-0.1599*
(0.0796)
Main
Treatment
0.0204
(0.0112)
-0.0105
(0.0110)
Note: * p<0.05, ** p<0.01. Standard errors are in parentheses.
Our analysis shows that being reassigned to an MCD lowered PRI vote share and
increased electoral competition in an election that took place three years after the
redistricting reform. We acknowledge that this particular effect of MCDs may have
attenuated over time, as the PRI eventually made appeals to indigenous voters that were
competitive with opposition-party appeals. Even if this occurred, however, we would
argue that the short-term effects of MCDs had long-term implications for electoral
competition by opening space for opposition parties to penetrate long-time PRI
strongholds and be competitive in future elections.
In sum, our findings consistently show that MCDs in Mexico did not have an
effect on minority electoral participation but did increase electoral competition in the
election following redistricting. Being reassigned to an MCD did not appear to bring
Gerrymandering Opposition
28
more minority voters to the polls and perhaps demobilized majority-group voters. These
findings dispute the purported motive of MCD creation, increasing minority participation.
Furthermore, we consistently find a significant negative effect of MCDs on PRI vote
share in highly indigenous municipalities; indigenous voters reassigned to MCDs voted
less for the former dominant party than other indigenous voters. This finding is bolstered
by the negative effect of MCDs on vote margins; elections became more closely
contested as a result of MCDs. The main takeaway from our empirical analysis is that the
creation of MCDs had an effect on for whom indigenous voters voted, rather than
whether they turned out to vote.
Ecological Inference
Given that municipalities are composed of both indigenous and mestizo voters we
must verify that observed differences in outcomes between treatment and control
municipalities were in fact due to changes in indigenous voting rather than changes in
mestizo voting. We estimated the values of turnout and PRI vote share for indigenous and
mestizo populations in all 2,048 municipalities using Ecological Inference (EI) software
that implements hierarchical modeling with bootstrapping for standard errors (Alimadhi,
Bhaskar, Lau, & Wittenberg, 2007).
Gerrymandering Opposition
Table 3
Ecological Inference Estimates
All Districts
Treatment Group
(%)
(%)
2009 Outcomes
Voter Turnout
47
47
Indigenous
[46,49]
[42,52]
47
39
Non-Indigenous
[46,47]
[36,41]
PRI Vote share
49
41
Indigenous
[47,50]
[37,46]
42
48
Non-Indigenous
[41,43]
[45,51]
2003 Outcomes
Voter Turnout
44
43
Indigenous
[43,45]
[39,47]
42
39
Non-Indigenous
[41,43]
[36,42]
PRI Vote share
49
48
Indigenous
[47,50]
[43,53]
46
50
Non-Indigenous
[46,47]
[46,54]
29
Control Group
(%)
48
[46,49]
47
[46,48]
49
[48,51]
42
[41,42]
44
[43,46]
42
[42,43]
49
[48,51]
46
[45,47]
Note: Full Dataset N=2,048; Treatment Group N=135; Control Group N=1,914. 95% Confidence Intervals
are in brackets.
Table 3 reports estimates of voter turnout and PRI vote share among indigenous
and the full dataset, treatment municipalities, and control municipalities. In 2009, an
estimated 41 percent of the indigenous population reassigned to MCDs voted for the PRI,
while 49 percent of the indigenous population in control units voted for the PRI. EI
estimates also report that in 2003, 48 percent of indigenous voters in the treatment group
voted for the PRI, compared to 49 percent of indigenous voters in the control group. This
supports our argument that indigenous voters newly reassigned to MCDs voted less for
the PRI (48 percent in 2003 versus 41 percent in 2009), while support for the PRI among
Gerrymandering Opposition
30
indigenous voters who were not reassigned to MCDs was unchanged (49 percent in both
years).15 EI estimates of indigenous turnout increased by roughly the same amount from
2003 to 2009 in treatment and control municipalities, supporting our finding that there
was no significant effect of MCDs on indigenous voter turnout.16 Notably, estimates for
indigenous and non-indigenous turnout in the pre-treatment (2003) full sample are quite
similar and, if anything, indigenous voters appear to have turned out in larger numbers
than mestizo voters, suggesting that minority electoral participation had not been
suppressed in the pre-reform context.
Opposition-Party Mobilization: Evidence of the Mechanism from Chiapas
Here, we analyze field research findings and secondary sources from the state of
Chiapas to illustrate that the mechanism explaining increased electoral competition in
MCDs was the opposition party’s adoption of an electoral strategy designed to draw
indigenous voters from the PRI. Prior to the appearance of the PRD in this state,
indigenous communities had turned out in large numbers for the PRI, owing to the
dominant party’s patronage appeals and top-down control, but were not afforded spaces
for self-government, either within or outside the party (Fox, 1996; Sonnleitner, 2012a). In
the years immediately prior to the adoption of MCDs, the Chiapas PRD forged alliances
15
We also observe greater vote shares for the PRI and lower turnout for mestizo voters
reassigned to MCDs than mestizo voters in control municipalities, suggesting that this
treatment demobilized this group and caused it to support opposition parties less.
16 We
also implemented an original field-research based EI technique, described in
greater detail in Meng & Palmer-Rubin (2012). Gerrymandering Opposition
31
with leaders of indigenous organizations, offering these organizations candidacies and
patronage rewards. Because its structure in Chiapas was weighted toward indigenous
communities, the PRD was poised to be more competitive in the newly created MCDs.
PRI leaders were unable to offer the same level of inducements due to their longstanding
commitments to mestizo elites. Our findings—an increase in electoral competition
accompanied by no change in indigenous turnout—are explained by the fact that the PRD
mobilized indigenous voters by wooing local leaders who had previously been aligned
with the PRI.
The left-wing PRD emerged on the electoral scene in Chiapas with the successful
gubernatorial campaign of Pablo Salazar Mendiguchía in 2000. With Salazar’s backing,
the state PRD operation embarked on a strategy to attract indigenous support by
nominating local indigenous leaders for elected office and government posts and
augmenting investments in indigenous communities. The PRD penetrated these
communities by forming alliances with leaders of indigenous associations, who often
wield substantial power in their communities as brokers in patronage networks and
kingmakers in local elections. These organizations—such as ARIC, CIOAC, and
UNORCA—originated in land invasions in the 1970s and were reinvigorated during the
1994 Zapatista movement.17 Many of these leaders, who had previously supported the
PRI given the lack of a viable alternative, were eager to join forces with an opposition
party that was electorally competitive and tolerant of their autonomous organizational
17
The organization that spearheaded the Zapatista rebellion, the Ejercito Zapatista de
Liberación Nacional (National Zapatista Liberation Army) withheld from forming party
alliances.
Gerrymandering Opposition
32
structures. Deviating to the PRD was most desirable for indigenous organizations in
MCDs, where indigenous leaders were relatively certain that this party could topple the
PRI if it garnered widespread support among indigenous voters.
These alliances also offered economic incentives for indigenous communities.
The organizations became channels through which the PRD provided patronage benefits,
including agricultural subsidies, social programs, and infrastructure investments, which
often exceeded the inducements that these communities had received from PRI
administrations. For example, an interviewed leader of CIOAC in the municipality of
Comitán explained that he was drawn to the PRD because the Salazar administration
channeled dozens of housing subsidies to its members, a benefit that had not been
afforded under the PRI (Antonio Hernández Cruz, personal communication, July 5,
2012). The PRD also granted nominations for elected office and government posts to
indigenous leaders. One of the main attractions of aligning with the PRD was the
guarantee of a nomination for mayor. Controlling municipal government availed
indigenous organizations of patronage resources, such as job posts and discretionary
infrastructure spending. Through these alliances, a number of indigenous organization
leaders also ran for state and federal congressional posts and occupied positions in the
state government and in the PRD’s state-level organization.18
18
Analysis by Sonnleitner (2012b, 91-115) shows that indigenous nominations to the
federal Chamber of Deputies increased significantly following the redistricting. In the
two federal elections prior to redistricting, four indigenous candidates were elected to this
chamber in 2000 and seven were elected in 2003. Following redistricting, 18 were elected
in 2006 and 16 were elected in 2009. However, our own analysis did not find that
Gerrymandering Opposition
33
During congressional elections, PRD-affiliated indigenous organizations led
campaigns on behalf of PRD candidates, indigenous or mestizo. In some municipalities,
the organizations assumed leadership of the party. For example, in the municipality of
Chilón, which is 96 percent indigenous, the PRD broke the PRI’s electoral monopoly by
allying with an indigenous organization called Yomlej and granting it power over the
candidate nomination process (Palmer-Rubin, 2011). The PRI in Chilón, on the other
hand, was dominated by mestizos in the PRI-affiliated peasant and ranchers’ associations
and thus was resistant to incorporating indigenous leaders into the party (Bobrow-Strain,
2007). Referring to another highly indigenous municipality in Chiapas, an indigenous
leader explained: “In the case of Las Margaritas, the PRD’s main force is the CIOAC.
When people think of the PRD, they think of the CIOAC” (Margarito Ruíz Hernández,
personal communication, July 4, 2012). Leaders described these arrangements as
providing an opportunity for indigenous communities to exercise “autonomous selfgovernment” and they returned the favor by organizing local campaigns for PRD
candidates at other levels of government. As a result, districts with large indigenous
populations offered the PRD its best opportunity to beat the PRI. For example, Chiapas’
second district became a PRI-PRD battleground after the 2005 redistricting caused the
district’s indigenous population to increase from 14 percent to 76 percent. After having
indigenous nominees were more common in newly formed MCDs than in previous
MCDs. This accords with qualitative findings; indigenous organization and party
representatives reported that nominations of local authorities to municipal office was a
much more common inducement offered to indigenous groups than legislative
nominations.
Gerrymandering Opposition
34
lost the 2003 congressional election in this district by a 16-percentage point margin, the
PRD edged out all other parties in 2009.
Given the increased salience of the indigenous vote, why wasn’t the PRI able to
augment its appeals to indigenous voters to counteract the PRD’s strategy? Despite the
PRI’s impressive patronage network in indigenous communities, the once-dominant
party’s organization in the Chiapas countryside, as elsewhere in Mexico, is entrenched in
the Central Nacional Campesina (National Peasant Confederation, CNC), a mestizodominated peasant confederation dating back to the 1930s. The PRI was unable to offer
the same level of inducements to indigenous leaders as the PRD because it was wedded
to mestizo elites. For example, one indigenous member of the PRI’s municipal council in
San Juán Chamula, regarded as a PRI bastion, said that even though the municipality is
close to 100 percent indigenous, the PRI’s state operation has resisted granting leadership
posts or candidacies to indigenous leaders: “After the [1994] resurrection we realized that
the PRI government had abandoned indigenous pueblos. It did not give us the opportunity
to be leaders or candidates, be it at the federal, state, or municipal level. We have been
asking for our fair share, which the party statutes provide for. But we have had to fight
for it” (Gilberto Velasco Rodríguez, personal communication, July 9, 2012). This
embattled operative also affirmed that indigenous communities were disfavored in the
distribution of party-mediated resources because PRI elites allocate a large share of
distributive programs to the mestizo-dominated CNC: “Indigenous pueblos participate
very little in the CNC, because there are other interests there. Very strong interests of
farmers and ranchers that disguise themselves as peasants, when in reality all of the
Gerrymandering Opposition
35
resources that go to the CNC are for certain groups in power, but very little actually ends
up going to projects for indigenous communities.”
To conclude, the 2005 redistricting took place in a context where the PRD in
Chiapas was gaining ground among indigenous communities by forging alliances with
local indigenous leaders, many of whom had previously been embedded in the PRI’s
patronage network. This strategy was more successful in MCDs than elsewhere because
these districts encompassed newly influential indigenous voting blocs, which were
amenable to opposition-party appeals and saw that the PRD stood a strong chance of
winning if it commanded the support of key indigenous leaders. The defection of these
leaders to the PRD explains why the creation of MCDs led to a decrease in support for
the PRI. The fact that this new electoral base previously supported the PRI, rather than
abstaining from electoral politics altogether, explains why MCDs had no discernable
effect on indigenous turnout.
Conclusion
This paper aimed to make two contributions to the study of minority-promoting
electoral institutions in new democracies. First, we sought to extend the analysis of
MCDs outside the United States context, testing whether these institutions had the
intended effect of bolstering minority electoral participation in Mexico. Using an original
dataset and a matching design that helps alleviate causal inference problems inherent to
observational studies, we find no effect of MCDs on minority voter turnout. Second, we
analyzed the impact of MCDs on electoral competition and one-party dominance during a
period in which Mexico was still emerging from the shackles of dominant-party rule.
Gerrymandering Opposition
36
We argued that MCDs promoted electoral competition by creating newly
influential minority voting blocs, which opposition parties were able to mobilize to
penetrate dominant-party strongholds. Fieldwork-based findings from the highly
indigenous state of Chiapas provided evidence of this mechanism. The PRD capitalized
on MCD creation to build local party organizations around indigenous organizations by
nominating their leaders to elected office and offering economic inducements. Prior to
the appearance of the PRD in Chiapas, indigenous voters had largely supported the PRI,
despite being politically marginalized. Restricted by ties to mestizo constituencies, the
PRI was unable to match the benefits that the PRD offered to indigenous communities.
We estimated the effects of MCDs using an identification strategy that exploited
variation induced by Mexico’s 2005 redistricting reform. We found evidence that MCDs
led to narrower vote margins and lower vote shares for the dominant party in highly
indigenous municipalities, yet no indication that they bolstered indigenous turnout. In
contrast with most extant research on the effect of minority-promoting electoral
institutions, we conclude that MCDs influenced for whom indigenous voters were voting,
not whether they turned out to vote.
These findings have important implications for minority representation in Mexico.
Since MCDs did not affect indigenous turnout, it appears that they did not “empower”
minority groups in the way that the literature on majority-minority districts in the United
States posits. Nonetheless, indigenous voters in Mexico benefited from the reform in a
different way. MCDs created incentives for opposition parties to offer greater
inducements to indigenous communities than they had received under the dominant party.
On the other hand, it appears unlikely that this reform substantially improved indigenous
Gerrymandering Opposition
37
representation on a national scale, as the most common inducements were mayoral
nominations and patronage benefits. Further research on MCDs in Mexico should assess
whether these districts led to more indigenous candidates being elected to congress
(“descriptive” representation) or more legislative attention given to indigenous interests
(“substantive” representation), outcomes that we were unable to observe given data
limitations.
Our study also contributes to the comparative politics literature more broadly.
First, we have shown that it is essential to understand pre-existing alignments between
ethnic minorities and political parties to be able to evaluate the causal effect of MCDs on
minority electoral participation. MCDs did not cause greater indigenous turnout in
Mexico because indigenous voters were already voting for the dominant party, albeit
without enjoying much clout in the party. Compared with other Latin American countries
with significant indigenous populations—such as Bolivia, Ecuador, Guatemala, or Peru –
Mexico is unique in the degree to which these populations were mobilized electorally by
a political party that pre-dated late-20th century indigenous movements. Thus, we would
expect that the creation of MCDs in these other countries would be more likely to
generate an empowerment effect of increasing minority turnout as found in the United
States.
Second, this paper makes a case that MCDs can contribute to the decline of oneparty dominance, one of the most trenchant obstacles to democratic consolidation. We
provide evidence that the creation of institutions that concentrate minority groups into
sizable voting blocs can open a window of opportunity for opposition parties to build new
constituencies. Party switching among minority voters in new democracies is a surprising
Gerrymandering Opposition
38
consequence of MCDs that has not been explored in the literature, likely due to the fact
that theories regarding MCDs are based on the contemporary United States, where most
minorities are loyal to the Democratic Party. Our findings show that MCD creation
improved the prospects of opposition parties in Mexico, a country with a history of oneparty dominance and top-down incorporation of ethnic minorities.
Thus, we expect that the effect of MCDs on electoral competition, at least through
the mechanism that we identify, is generalizable to former dominant-party regimes. In
addition to Mexico, the mobilization of marginalized minorities by opposition parties has
been shown to introduce electoral competition in the former Democrat-controlled US
South and in India, which was dominated by the Congress Party. Conversely, other
highly indigenous countries in Latin America passed through military dictatorships that
excluded or violently repressed the indigenous rather than one-party dominant regimes
that incorporated them into patronage networks. We expect that the type of party systems
that emerged in these countries after their transitions to democracy and the ability of
parties to incorporate indigenous voters would modify the effect of MCDs. Future
research may investigate how minority-promoting institutions affect electoral competition
in newly democratic countries without a history of one-party dominance.
Gerrymandering Opposition
39
References
Alimadhi F, Bhaskar BN, Lau O, Wittenberg J. ei.RxC: Hierarchical multinomialdirichlet ecological inference model for R x C tables; 2007.
Ansolabehere S, Snyder JM, Stewart C. Old voters, new voters, and the personal vote:
Using redistricting to measure the incumbency advantage. Am J Polit Sci
2000:44(1):17-34.
Barreto M. Si se puede! Latino candidates and the mobilization of Latino
voters. Am Polit Sci Rev 2007;101(3):425-441.
Barreto M, Villarreal M, Woods N. Metropolitan Latino political behavior: Voter turnout
and candidate preference in Los Angeles. J Urban Aff 2005;27(1):71-91.
Benton A. Bottom-up challenges to national democracy: Mexico’s (legal) subnational
authoritarian enclaves. Comp Polit 2012;44(3):253-271.
Birnir JK. Divergence in diversity? The dissimilar effects of cleavages on electoral
politics in new democracies. Am J Polit Sci 2007;51(3):602-619.
Bobrow-Strain A. Intimate enemies: Landowners, power, and violence in Chiapas.
Durham: Duke University Press; 2007.
Bobo L, Gilliam Jr. FD. Race, sociopolitical participation, and black empowerment. Am
Polit Sci Rev 1990;84(2):377-93.
Brace K, Grofman B, Handley L. Does redistricting aimed to help blacks necessarily help
Republicans? J Polit 1987;49(1):169–185.
Cameron C, Epstein D, O’Halloran S. Do majority-minority districts maximize
substantive black representation in congress? Am Polit Sci Rev 1996;90(4):794812.
Gerrymandering Opposition
40
Chandra K. Why ethnic parties succeed: Patronage and ethnic head counts in India. New
York: Cambridge University Press; 2004.
Chhibber P. Democracy without associations: Transformation of the party system and
social cleavages in India. Ann Arbor: University of Michigan Press; 1999.
Cleary M. Indigenous autonomy and insulation from electoral competition in Oaxaca,
Mexico. Paper presented at the conference on Native Peoples in Mexico and the
U.S.: approaches to exclusion, San Diego, CA; 2009.
Consejo Nacional de Evaluación de la Política de Desarrollo Social (CONEVAL). 2010.
Análisis y medición de la pobreza: anexo estadístico - pobreza 2010. Retrieved
from: <http://www.coneval.gob.mx/Medicion/Paginas/Medici%C3%B3n/AnexoEstadistico-Pobreza-2010.aspx>
Diaz-Cayeros A, Magaloni B, Weingast BR. Tragic brilliance: Equilibrium hegemony
and democratization in Mexico. Working Paper, Hoover Institution, Stanford
University; 2003.
Eisenstadt T. Usos y costumbres and postelectoral conflicts in Oaxaca, Mexico, 19952004: An empirical and normative assessment. Lat Am Res Rev 2007;42(1): 52–
77.
Fox J. How does civil society thicken? The political construction of social capital in rural
Mexico. World Dev 1996;24(6):1089–1103.
Gay C. The effect of black congressional representation on political participation. Public
Policy Institute of California. 2001; 95(3), 589-602.
Gibson E. Boundary control: Subnational authoritarianism in federal democracies.
Cambridge: Cambridge University Press; 2012.
Gerrymandering Opposition
41
Giraudy A. The politics of subnational undemocratic regime reproduction in Argentina
and Mexico. J Polit Lat Am 2010;2(2):54-84.
Goodnow R, Moser R. Layers of ethnicity: The effects of ethnic federalism, majorityminority districts, and minority concentration on the electoral success of ethnic
minorities in Russia. Comp Polit Stud 2012;45(2):167-193.
Harvey N. The Chiapas rebellion: The struggle for land and democracy. Durham: Duke
University Press; 1998.
Hayes D, McKee SC. The participatory effects of redistricting. Am J Polit Sci 2009;
53(4):1006-1023.
Hayes D, McKee SC. The intersection of redistricting, race, and participation. Am J Polit
Sci 2012;56(1):115-130.
Henderson J, Sekhon J, Titiunik R. Cause or effect? Turnout in Hispanic majorityminority districts. Working Paper; 2012.
Herron MC, Sekhon J. Black candidates and black voters: Assessing the impact of
candidate race on uncounted vote rates. J Polit 2005;67(1):154–77.
Hill KA. Does the creation of majority black districts aid Republicans? An analysis of the
1992 congressional elections in eight southern states. J Polit 1995;57(2):384–401.
Instituto Federal Electoral (IFE). Acuerdo del Consejo General del Instituto Federal
Electoral por el que se establece la demarcación territorial de los trescientos
distritos electorales federales uninominales en que se divide el país para su
utilización en los procesos electorales federales 2005-2006 y 2008-2009,
Agreement CG28/2005; 2005a.
Instituto Federal Electoral (IFE). Distritación 2004-2005: camino para la democracia;
Gerrymandering Opposition
42
2005b.
Jensenius F. Power, performance, and bias: Evaluating the electoral quotas for scheduled
castes in India. (Doctoral dissertation). University of California, Berkeley; 2013.
Keele L, White I. African American turnout in majority-minority districts. Paper
presented at the Annual Meeting of the American Political Science Association,
Seattle, WA; 2011.
Laakso M, Taagepera R. ‘Effective’ number of parties: A measure with application to
West Europe. Comp Polit Stud 1979;12(1):3-27.
McMillan A. Standing at the Margins: Representation and Electoral Reservation in India.
New Delhi: Oxford University Press; 2005.
Mickey R. The beginning of the end for authoritarian rule in America:
Smith v. Allwright and the abolition of the white primary in the Deep South,
1944-1948. Stud Am Polit Dev 2008;22(2):143-182.
Author. From frame to campaign: Indigeneity in the political arena. Paper presented at
the Annual Meeting of the American Political Science Association, Seattle, WA;
2011.
Posner D. Institutions and ethnic politics in Africa. New York: Cambridge University
Press; 2005.
United Nations Development Program (UNDP). Informe sobre desarrollo humano de los
pueblos indígenas en México: El reto de la desigualdad de oportunidades; 2010.
Rus J. The “comunidad revolucionaria institucional:” The subversion of native
government in highland Chiapas. In: Joseph G, Nugent D, editors. Everyday
forms of state formation: Revolution and the negotiation of rule in modern
Gerrymandering Opposition
43
Mexico. Durham: Duke University Press; 1994. p.265-300.
Sekhon J. Multivariate and propensity score matching software with automated balance
optimization: The matching package for R. J Stat Softw 2011;42(7):1-52.
Sekhon J, Titiunik R. When natural experiments are neither natural nor experiments. Am
Polit Sci Rev 2012;106(1):35-57.
Sonnleitner W. Elecciones chiapanecas: Del régimen posrevolucionario al desorden
democrático. Mexico City: El Colegio de México; 2012a.
Sonnleitner W. La representación legislativa de los indígenas en México: De la
representatividad descriptiva, a una representación de mejor calidad; 2012b.
Tate K. Black political participation in the 1984 and 1988 presidential elections. Am Polit
Sci Rev 1991;85(4):1159-1176.
Trejo G. Religious competition and ethnic mobilization in Latin America: Why the
Catholic church promotes indigenous movements in Mexico. Am Polit Sci Rev
2009;103(3):323–342.
Valelly R. The two reconstructions: The struggle for black enfranchisement. Chicago:
University of Chicago Press; 2004.
Voss S, Lublin D. Black incumbents, white districts: An appraisal of the 1996
congressional elections. Am Polit Res 2001;29(2):141-182.
Gerrymandering Opposition
44
Appendices
Appendix A
Summary Statistics, Full Dataset
Old District % Indigenous
New District % Indigenous
Municipal % Indigenous
Municipal Spending per
capita
Municipal % Households
with Concrete Floors
Municipal Population
2009 turnout
2009 PRI vote share
2009 margin of victory
2003 turnout
2003 PRI vote share
2003 margin of victory
1997 turnout
1997 PRI vote share
1997 margin of victory
Note: N=2,048.
Mean
(Standard Deviation)
0.21
(0.23)
0.22
(0.25)
0.22
(0.33)
2,006.24
(1,151.05)
0.72
(0.23)
33,478
(152,554)
0.47
(0.14)
0.43
(0.13)
0.20
(0.15)
0.42
(0.13)
0.47
(0.13)
0.21
(0.17)
0.53
(0.12)
0.52
(0.14)
0.26
(0.20)
[Min, Max]
[0, 0.75]
[0, 0.78]
[0, 1]
[138.38, 16,579]
[0.04, 0.99]
[102, 5,368,545]
[0.09, 0.85]
[0.07, 0.90]
[0, 0.87]
[0, 0.99]
[0, 0.99]
[0, 0.97]
[0.04, 0.99]
[0.15, 1]
[0, 0.99]
Gerrymandering Opposition
45
Appendix B
Criteria for 2005 Redistricting Law19
1. District lines must fall within states
2. Districts populations should be within 15% of average district size
a. If a municipality is too big, it can split into multiple districts (urban cases)
3. When states are smaller than average district size, ad-hoc decisions are acceptable
4. District boundaries should be drawn to preserve territorial integrity of indigenous
communities
5. Districts should be geographically and politically continuous (for example, rural
development regions should not be split)
6. Anti-Gerrymandering Clause: District boundaries should be as small as possible
and preferably octagonal; one district cannot completely surround another
7. The smallest unit for redistricting is the precinct
8. Where possible, municipalities should not divided into different districts
a. Suburbs should be included with the associated city
b. Urban localities are defined as 15,000 people
c. Rural and urban localities should not be mixed
9. The district capital should be a city with a large population, transportation access,
and good public services
10. When generating districts, travel time across districts should be made as short as
possible
19
Authors’ summary of IFE (2005b).
Gerrymandering Opposition
46
Appendix C
Pre- and Post-Matching Balance between Treatment and Control Groups
Mean
Treated
Mean
Control
Old.District.Percent.Indig
0.249
0.267
●
Municipality.Percent.Indig
0.393
0.377
●
Log.Spending.per.Capita
7.411
7.425
●●
Percent.Concrete.Floors
0.617
0.636
●
●
Percent.New.Neighbors
0.704
0.677
PRI.Vote.Share.2003
0.501
0.51
Turnout.2003
0.398
0.393
●
PRI.Vote.Share.1997
0.543
0.552
●●
Turnout.1997
0.48
0.495
BM KS p−value
AM KS p−value
●
●
1.00
●
●●
0.05
BM t p−value
AM t p−value
●
0.00
●
9643.317 11090.797
0.10
Population
P−values
Note: Reported means correspond to post-matching treatment and control groups. Placement of circles and
triangles represent t-test and Kolmogorov-Smirnov (KS) test probabilities for differences between prematching and post-matching treatment and control groups. Blue shapes correspond to t-test and red shapes
correspond to KS test.
Table 1: Control and Treatment Units by State and Governing Party
Ruling Party
Ruling Party
Control
in 2003
in 2009
Units
Chiapas
PRD
PRD
8
Chihuahua
PRI
PRI
2
Durango
PRI
PRI
3
Guerrero
PRI
PRD
3
Hidalgo
PRI
PRI
4
Mexico
PRI
PRI
3
Treatment
Units
16
0
0
6
1
0
Gerrymandering Opposition
Michoacán
Nayarit
Oaxaca
Puebla
Veracruz
Yucatan
Total
PRD
PAN
PRI
PRI
PRI
PAN
47
PRD
PRI
PRI
PRI
PRI
PRI
7
2
59
26
6
0
123
0
0
67
20
6
7
123
Table 2: Summary of Control and Treatment Units by State Governing Party
PRI
PRD
PAN
Control
106
15
2
2003
Treatment
100
16
7
Control
105
18
0
2009
Treatment
101
22
0
Table 1 shows the number of treatment and control units in each state in the
analysis and also lists the party of the governor of each state in 2003 and 2009 as an
indicator of party strength. MCDs were not created in all states, and in some states where
MCDs were created, they encompassed all of the majority indigenous municipalities
(leaving no control units) or treatment units had no good matches on covariates. To
prioritize the quality of the matches, we chose not to restrict matching within states.
However, findings are robust to within-state matching where available, but this results in
extreme imbalance on several covariates and requires the exclusion of several treated
municipalities. Importantly, we matched treatment and control municipalities on 1997
PRI vote share, assuring that pre-treatment voting patterns were consistent across control
and treatment municipalities, regardless of state-level trends in party strength.
Oaxaca and Puebla are the two states that have by far the most units in the
analysis. This is due to the fact that these are highly fragmented states (they have 570 and
217 total municipalities, respectively) and that they also have significant indigenous
populations. Both of these states were PRI dominant up until the 2010 election, when
candidates running with support from a PAN-PRD coalition unseated the PRI. Given that
there are roughly equal treatment and control units in each state, the effect of this
electoral breakthrough in these two states should not fall disproportionately on either
treatment or control units in the analysis.
Table 2 summarizes Table 1 showing how many treatment and control units were
in states governed by each of the parties in 2003 and 2009. The number of both types of
units in PRI-governed states is quite similar in both 2003 (106 control, 100 treatment)
and 2009 (105 control, 101 treatment) as are the number of municipalities in PRDgoverned states.
Gerrymandering Opposition
48
Appendix D
Description of Datasets and Treatments
Dataset
Description
Full dataset
Full matched dataset
Majority indigenous
matched dataset
All municipalities
All matched treatment and
control municipalities from
the full dataset
All majority indigenous
treatment municipalities (at
least 50% indigenous) from
full matched dataset with
corresponding control
municipalities
Treatment Name
Main Treatment
Continuous Treatment
Magnitude Effect
Threshold Effect
Reassignment Effect
Units
N=2,048
Treat=135
Control=1,914
N=246
Treat=123
Control=123
N=88
Treat=44
Control=44
Description
50% increase in the percent indigenous
from the old district to the new district and
the new district is at least 40% indigenous
Percent indigenous of the municipality’s
new district minus the percent indigenous
of the municipality’s old district
50% increase in the percent indigenous
from the old district to the new district
Old district was less than 40% indigenous
and new district is more than 40%
indigenous
Treated units are municipalities that were
newly assigned from districts below 40%
indigenous to MCDs. Control units are
municipalities that have always been in
MCDs.
Gerrymandering Opposition
49
Appendix E
Sensitivity Analysis Estimates using Alternate Treatment Conditions
Outcome: voter
Outcome: PRI vote
Outcome: vote
turnout
share
margin
1997
1997
1997
Placebo
2009
Placebo
2009
Placebo
2009
Test
Test
Test
Magnitude
Effect
-.0257
(.0191)
.0126
(.0160)
-.0095
(.0198)
-.0547*
(.0199)
.0021
(.0294)
-.0785*
(.0251)
Threshold
Effect
Reassignment
Effect
-.0126
(.0192)
.0032
(.0158)
.0117
(.0160)
.0134
(.0167)
-.0386
(.0196)
.0035
(.0170)
-.0703*
(.0197)
-.0362
(.0151)
-.0482
(.0291)
.0001
(.0251)
-.1002**
(.0246)
-.0486
(.0194)
Note: * p<0.05, ** p<0.01. Standard errors are in parentheses.
Magnitude Effect: N = 88. Analysis includes 48 treatment and 40 control units.
Estimates calculated using all majority indigenous treatment units from original matched
dataset with corresponding controls. Findings discussed in text.
Threshold Effect: N = 88. Analysis includes 49 treatment and 39 control units.
Estimates calculated using all majority indigenous treatment units from original matched
dataset with corresponding controls. Finding discussed in text.
Reassignment Effect: N = 136. Analysis includes 68 treatment and 68 control units.
Estimates calculated using all majority indigenous treatment districts and corresponding
controls from a matched dataset constructed from all municipalities whose new district is
greater than 40 percent indigenous. Treatment is defined as being newly in an MCD—old
district was less than 40 percent indigenous. Control units are those whose old and new
districts are both greater than 40 percent indigenous.
This test allows us to adjudicate between the effect of simply being in an MCD
and the effect of having recently been reassigned to an MCD. We find no effect of
reassignment on turnout and a negative effect on PRI vote share and vote shares, just
short of statistical significance given the standard 95 percent confidence levels. We must
interpret these findings with caution because there is a high risk of confounding given
that treatment and control units were in different types of districts prior to redistricting.
However, the contrast in voting patterns between municipalities newly in MCDs and
municipalities that have always been in MCDs suggests that it was not merely the
contextual effect of highly indigenous districts that caused the PRI to lose ground, but
particularly the effect of newly created indigenous voting blocs.